Lucid / api_server.py
BigJuls's picture
we cool bruh
87d8d6b
Raw History Blame
30.5 kB
"""
FastAPI Server for Lucid 2.0 Surveillance System
Provides real-time camera streaming and AI integration
"""
import cv2
import numpy as np
import asyncio
import threading
import time
import os
from typing import Generator, List, Tuple, Dict, Optional
from fastapi import FastAPI, HTTPException, Response, WebSocket, WebSocketDisconnect, UploadFile, File
from fastapi.responses import StreamingResponse, FileResponse
from fastapi.middleware.cors import CORSMiddleware
import uvicorn
import json
from datetime import datetime
# Import existing components
from advanced_camera_manager import AdvancedCameraManager
from enhanced_tracking_system import EnhancedTrackingSystem
from realtime_embedding_matcher import RealtimeEmbeddingMatcher
app = FastAPI(title="Lucid 2.0 API", version="1.0.0")
# Add CORS middleware
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# Global instances
camera_manager = None
tracking_system = None
matcher = None
# WebSocket connections
active_connections = []
# Person profiles storage
person_profiles = {} # {global_id: {name, photo, metadata}}
# Frame cache for streaming
frame_cache = {}
frame_cache_lock = threading.Lock()
# Alert rules storage
alert_rules = [] # List of alert rule configurations
def initialize_system():
"""Initialize the surveillance system components"""
global camera_manager, tracking_system, matcher
print("Initializing Lucid 2.0 system...")
# Initialize camera manager
camera_manager = AdvancedCameraManager()
# Initialize tracking system
tracking_system = EnhancedTrackingSystem()
if not tracking_system.initialize_system():
raise Exception("Failed to initialize tracking system")
# Initialize matcher
matcher = RealtimeEmbeddingMatcher()
matcher.start_matching()
print("✓ Lucid 2.0 system initialized")
async def broadcast_detection(data):
"""Broadcast detection data to all active WebSocket connections"""
disconnected = []
for connection in active_connections:
try:
await connection.send_json(data)
except WebSocketDisconnect:
disconnected.append(connection)
except Exception as e:
print(f"Error sending to WebSocket: {e}")
disconnected.append(connection)
# Remove disconnected clients
for connection in disconnected:
if connection in active_connections:
active_connections.remove(connection)
async def broadcast_alert(data):
"""Broadcast alert data to all active WebSocket connections"""
disconnected = []
for connection in active_connections:
try:
await connection.send_json(data)
except WebSocketDisconnect:
disconnected.append(connection)
except Exception as e:
print(f"Error sending alert to WebSocket: {e}")
disconnected.append(connection)
# Remove disconnected clients
for connection in disconnected:
if connection in active_connections:
active_connections.remove(connection)
def check_alert_rules(identity_data):
"""Check if any alert rules are triggered"""
global alert_rules, active_connections
if not alert_rules:
return
print(f"[Alert Rules] Checking {len(alert_rules)} rules for identity {identity_data.get('global_id', 'unknown')}")
for rule in alert_rules:
rule_type = rule.get("type")
triggered = False
alert_message = None
# 1. Person Reappearance
if rule_type == "reappearance":
if rule.get("person_id") == identity_data.get("global_id"):
triggered = True
alert_message = f"Person {rule.get('person_name', identity_data.get('global_id')[:12])} has reappeared"
# 2. Unknown Person After Hours
elif rule_type == "unknown_after_hours":
current_hour = datetime.now().hour
if current_hour >= 21 or current_hour < 6: # After 9 PM or before 6 AM
if identity_data.get("confidence", 0) < 0.8: # Unknown/low confidence
triggered = True
alert_message = f"Unknown person detected after hours (confidence: {identity_data.get('confidence', 0):.2f})"
# 3. High Confidence Detection
elif rule_type == "high_confidence":
threshold = rule.get("threshold", 0.8)
if identity_data.get("confidence", 0) >= threshold:
triggered = True
alert_message = f"High confidence detection: {identity_data.get('confidence', 0):.2f}"
# 4. Loitering Detection (person seen for extended time)
elif rule_type == "loitering":
# This would need dwell time calculation - placeholder for now
cameras_seen = identity_data.get("cameras_seen", [])
if len(cameras_seen) >= 1: # Detected in at least one camera
triggered = True
alert_message = f"Person detected in {len(cameras_seen)} camera(s)"
# Send alert if triggered
if triggered and alert_message:
alert_data = {
"type": "alert",
"alert_type": rule_type,
"message": alert_message,
"timestamp": time.time(),
"data": identity_data
}
print(f"[Alert] Rule triggered: {rule_type} - {alert_message}")
# Broadcast to all connected clients
asyncio.run_coroutine_threadsafe(
broadcast_alert(alert_data),
asyncio.get_event_loop()
)
def capture_frames_continuously(camera_id: str):
"""Continuously capture frames and store in cache for streaming"""
global camera_manager, frame_cache, tracking_system, matcher
print(f"[Streaming] Started frame capture for {camera_id}")
print(f"[DEBUG] tracking_system initialized: {tracking_system is not None}")
print(f"[DEBUG] camera_manager initialized: {camera_manager is not None}")
frame_count = 0
while True:
try:
# Check if components are initialized
if camera_manager is None or tracking_system is None:
time.sleep(1)
continue
# Get frame from camera manager
ret, frame = camera_manager.get_frame(camera_id)
print(f"[DEBUG] {camera_id}: Got frame - ret={ret}, frame shape={frame.shape if frame is not None else 'None'}")
if ret and frame is not None:
frame_count += 1
if frame_count % 30 == 1: # Log every 30 frames
print(f"[DEBUG] {camera_id}: Processing frame {frame_count}")
# Process frame through tracking system
frame_number = int(time.time() * 1000) % 1000000
detections = tracking_system.process_frame(frame, camera_id, frame_number)
if frame_count % 30 == 1:
print(f"[DEBUG] {camera_id}: Got {len(detections)} detections")
# Draw detections on frame
display_frame = frame.copy()
# Draw detections
for det in detections:
x1, y1, x2, y2 = det.bbox
# Draw bounding box
cv2.rectangle(display_frame, (x1, y1), (x2, y2), (0, 0, 255), 2)
# Draw label
label = f"Det: {det.id[:8]}"
if det.confidence:
label += f" ({det.confidence:.2f})"
cv2.putText(display_frame, label, (x1, y1 - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 255), 2)
# Draw tracks and create global identities
active_tracks = tracking_system.get_active_tracks(camera_id)
for track in active_tracks:
if track.detections:
latest_detection = track.detections[-1]
# FIXED: Ensure global ID is created for new tracks
if not track.global_id and track.representative_embedding is not None:
# Try cross-camera re-identification first
matched_global_id = tracking_system.cross_camera_reidentification(track)
if not matched_global_id:
# Create new global identity
tracking_system._create_global_identity(track)
if frame_count % 30 == 1:
print(f"[DEBUG] Created new global ID: {track.global_id} for track {track.id}")
# Check alert rules for new identity
if track.global_id:
identity = tracking_system.global_identities.get(track.global_id)
if identity:
identity_data = {
"global_id": track.global_id,
"confidence": identity.confidence,
"cameras_seen": identity.cameras_seen,
"person_name": track.global_id[:12]
}
check_alert_rules(identity_data)
x1, y1, x2, y2 = latest_detection.bbox
# Draw track box (green for tracked)
cv2.rectangle(display_frame, (x1, y1), (x2, y2), (0, 255, 0), 2)
# Draw track label with global ID
track_label = f"Track: {track.id}"
if track.global_id:
track_label += f" ({track.global_id})"
cv2.putText(display_frame, track_label, (x1, y2 + 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
# Store frame in cache
with frame_cache_lock:
frame_cache[camera_id] = display_frame
# Send real-time alerts for new detections
for det in detections:
if det.confidence > 0.7: # Only send high-confidence detections
detection_data = {
"type": "new_detection",
"camera_id": camera_id,
"detection_id": det.id,
"bbox": det.bbox,
"confidence": det.confidence,
"timestamp": time.time()
}
# Run in asyncio thread
asyncio.run_coroutine_threadsafe(
broadcast_detection(detection_data),
asyncio.get_event_loop()
)
# Control frame rate
time.sleep(0.033) # ~30 FPS
else:
time.sleep(0.1)
except Exception as e:
print(f"[Streaming] Error capturing from {camera_id}: {e}")
time.sleep(1)
def generate_frames(camera_id: str) -> Generator[bytes, None, None]:
"""Generate MJPEG frames for streaming"""
global frame_cache
print(f"[STREAM] Client connected to /video_feed/{camera_id}")
# Start capture thread if not already running
capture_thread = threading.Thread(
target=capture_frames_continuously,
args=(camera_id,),
daemon=True
)
capture_thread.start()
print(f"[STREAM] Started capture thread for {camera_id}")
while True:
try:
# Get frame from cache
with frame_cache_lock:
frame = frame_cache.get(camera_id)
if frame is not None:
# Encode frame as JPEG
ret, buffer = cv2.imencode('.jpg', frame, [cv2.IMWRITE_JPEG_QUALITY, 80])
if ret:
# Yield frame in MJPEG format
yield (b'--frame\r\n'
b'Content-Type: image/jpeg\r\n\r\n' + buffer.tobytes() + b'\r\n')
# Control stream rate
time.sleep(0.033) # ~30 FPS
except Exception as e:
print(f"[Streaming] Error generating frames for {camera_id}: {e}")
time.sleep(1)
@app.get("/")
async def root():
return {"message": "Lucid 2.0 Surveillance System API", "status": "online"}
@app.get("/api/cameras")
async def get_cameras():
"""Get list of all cameras"""
if camera_manager is None:
raise HTTPException(status_code=503, detail="Camera manager not initialized")
cameras = camera_manager.get_all_cameras()
return {
"cameras": [
{
"id": cam_id,
"source": cam_info.source,
"type": cam_info.type,
"status": cam_info.status,
"resolution": cam_info.resolution,
"fps": cam_info.fps
}
for cam_id, cam_info in cameras.items()
]
}
@app.post("/api/cameras")
async def add_camera(camera_id: str, source: str):
"""Add a new camera"""
if camera_manager is None:
raise HTTPException(status_code=503, detail="Camera manager not initialized")
success = camera_manager.add_camera(camera_id, source)
if success:
return {"message": f"Camera {camera_id} added successfully", "status": "success"}
else:
raise HTTPException(status_code=400, detail=f"Failed to add camera {camera_id}")
@app.delete("/api/cameras/{camera_id}")
async def remove_camera(camera_id: str):
"""Remove a camera"""
if camera_manager is None:
raise HTTPException(status_code=503, detail="Camera manager not initialized")
# Stop camera capture
camera_manager.remove_camera(camera_id)
return {"message": f"Camera {camera_id} removed", "status": "success"}
@app.get("/api/video_feed/{camera_id}")
async def video_feed(camera_id: str):
"""Stream video feed from a camera as MJPEG"""
if camera_manager is None:
raise HTTPException(status_code=503, detail="Camera manager not initialized")
# Check if camera exists
cameras = camera_manager.get_all_cameras()
if camera_id not in cameras:
raise HTTPException(status_code=404, detail=f"Camera {camera_id} not found")
# Return streaming response
return StreamingResponse(
generate_frames(camera_id),
media_type="multipart/x-mixed-replace; boundary=frame"
)
@app.get("/api/stats")
async def get_system_stats():
"""Get system statistics"""
if tracking_system is None or matcher is None:
raise HTTPException(status_code=503, detail="System not fully initialized")
tracking_stats = tracking_system.get_system_stats()
matcher_stats = matcher.get_system_statistics()
# Add global identity count
global_identity_count = len(tracking_system.global_identities)
return {
"tracking": tracking_stats,
"matcher": matcher_stats,
"global_identities": global_identity_count
}
@app.get("/api/identities")
async def get_identities():
"""Get all global identities with enhanced data"""
if tracking_system is None:
raise HTTPException(status_code=503, detail="Tracking system not initialized")
# FIXED: Get identities directly from tracking system instead of matcher
all_identities = tracking_system.global_identities
enhanced_identities = []
for global_id, identity in all_identities.items():
# Get profile if exists
profile = person_profiles.get(global_id, {})
# Create timeline of camera appearances from tracks
timeline = []
camera_timestamps = {} # {camera_id: [timestamps]}
for track in identity.tracks:
if track.camera_id not in camera_timestamps:
camera_timestamps[track.camera_id] = []
camera_timestamps[track.camera_id].append(track.first_seen)
timeline.append({
"camera_id": track.camera_id,
"first_seen": track.first_seen,
"last_seen": track.last_seen,
"detections": track.detection_count
})
# Sort timeline by time
timeline.sort(key=lambda x: x["first_seen"])
enhanced_identity = {
"global_id": identity.global_id,
"confidence": identity.confidence,
"cameras_seen": identity.cameras_seen,
"first_seen": identity.first_seen,
"last_seen": identity.last_seen,
"observations": len(identity.tracks),
"timeline": timeline,
"profile": profile if profile else None,
"snapshot_url": f"/snapshots/{global_id}.jpg" if os.path.exists(f"{tracking_system.snapshots_dir}/{global_id}.jpg") else None,
"heatmap": camera_timestamps # For heatmap visualization
}
enhanced_identities.append(enhanced_identity)
# Check alert rules
check_alert_rules(enhanced_identity)
print(f"[DEBUG] /identities endpoint returning {len(enhanced_identities)} identities")
return {
"identities": enhanced_identities
}
@app.post("/api/find_person")
async def find_person(image: UploadFile = File(...)):
"""Find a person across all cameras using Re-ID with improved accuracy"""
if tracking_system is None:
raise HTTPException(status_code=503, detail="System not fully initialized")
try:
# Read uploaded image
contents = await image.read()
nparr = np.frombuffer(contents, np.uint8)
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
if img is None:
raise HTTPException(status_code=400, detail="Invalid image file")
# Extract features from the uploaded image
face_emb, body_emb, _ = tracking_system.extract_features(img, (0, 0, img.shape[1], img.shape[0]))
if body_emb is None and face_emb is None:
raise HTTPException(status_code=400, detail="Could not extract features from image")
print(f"[Search] Extracted embeddings - Face: {face_emb is not None}, Body: {body_emb is not None}")
# Search against ALL global identities directly (more accurate than vector DB)
from scipy.spatial.distance import cosine
matches = []
for global_id, identity in tracking_system.global_identities.items():
best_similarity = 0.0
match_type = None
# Compare body embeddings
if body_emb is not None and len(identity.body_embeddings) > 0:
for body_emb_stored in identity.body_embeddings:
if body_emb_stored is not None:
similarity = 1 - cosine(body_emb, body_emb_stored)
if similarity > best_similarity:
best_similarity = similarity
match_type = "body"
# Compare face embeddings (more reliable, higher weight)
if face_emb is not None and len(identity.face_embeddings) > 0:
for face_emb_stored in identity.face_embeddings:
if face_emb_stored is not None:
similarity = 1 - cosine(face_emb, face_emb_stored)
# Face matching is more discriminative, so boost it slightly
if similarity > best_similarity:
best_similarity = similarity
match_type = "face"
# Only include matches above threshold (0.70 for balanced precision/recall)
if best_similarity >= 0.70:
# Get most recent appearance info
latest_track = max(identity.tracks, key=lambda t: t.last_seen) if identity.tracks else None
matches.append({
"global_id": global_id,
"confidence": float(best_similarity),
"match_type": match_type,
"camera_id": latest_track.camera_id if latest_track else "unknown",
"timestamp": identity.last_seen,
"first_seen": identity.first_seen,
"cameras_seen": identity.cameras_seen,
"observations": len(identity.tracks),
"entity_id": global_id # For compatibility
})
# Sort by confidence (highest first)
matches.sort(key=lambda x: x["confidence"], reverse=True)
print(f"[Search] Found {len(matches)} matches above threshold 0.70")
if len(matches) == 0:
print(f"[Search] No matches found. Total identities in system: {len(tracking_system.global_identities)}")
# Log similarity scores for debugging
for global_id, identity in list(tracking_system.global_identities.items())[:5]:
print(f" {global_id}: Face embeddings: {len(identity.face_embeddings)}, Body embeddings: {len(identity.body_embeddings)}")
return {
"matches": matches[:20], # Return top 20 matches
"total_matches": len(matches),
"search_method": "global_identity_direct"
}
except Exception as e:
print(f"Error in find_person: {e}")
import traceback
traceback.print_exc()
raise HTTPException(status_code=500, detail=f"Error processing image: {str(e)}")
@app.post("/api/profiles/{global_id}")
async def create_profile(global_id: str, name: str, metadata: Optional[str] = None):
"""Create or update a person profile"""
global person_profiles
profile = {
"name": name,
"created_at": time.time(),
"metadata": metadata or "{}"
}
person_profiles[global_id] = profile
return {
"message": f"Profile created for {global_id}",
"profile": profile
}
@app.get("/api/profiles/{global_id}")
async def get_profile(global_id: str):
"""Get a person's profile"""
if global_id in person_profiles:
return person_profiles[global_id]
else:
raise HTTPException(status_code=404, detail="Profile not found")
@app.post("/api/alerts/rules")
async def create_alert_rule(rule: dict):
"""Create a new alert rule"""
global alert_rules
alert_rules.append(rule)
return {"message": "Alert rule created", "rule": rule}
@app.get("/api/alerts/rules")
async def get_alert_rules():
"""Get all alert rules"""
return {"rules": alert_rules}
@app.get("/api/snapshots/{filename}")
async def get_snapshot(filename: str):
"""Serve snapshot images"""
if tracking_system is None:
raise HTTPException(status_code=503, detail="Tracking system not initialized")
snapshot_path = os.path.join(tracking_system.snapshots_dir, filename)
if not os.path.exists(snapshot_path):
raise HTTPException(status_code=404, detail="Snapshot not found")
return FileResponse(snapshot_path, media_type="image/jpeg")
@app.get("/api/analytics")
async def get_analytics():
"""Get comprehensive system analytics"""
if tracking_system is None:
raise HTTPException(status_code=503, detail="Tracking system not initialized")
current_time = time.time()
all_identities = tracking_system.global_identities
# 1. Overview Metrics
total_identities = len(all_identities)
active_identities = len([id for id, identity in all_identities.items()
if current_time - identity.last_seen < 300]) # Active in last 5 min
# 2. Camera Analytics
camera_stats = {}
cameras = camera_manager.get_all_cameras() if camera_manager else []
for camera in cameras:
camera_id = str(camera.get('id', '')) if isinstance(camera, dict) else camera
# Count unique people seen by this camera
unique_visitors = len([id for id, identity in all_identities.items()
if camera_id in identity.cameras_seen])
# Count observations in this camera
total_observations = sum([len([t for t in identity.tracks if t.camera_id == camera_id])
for identity in all_identities.values()])
camera_stats[camera_id] = {
"unique_visitors": unique_visitors,
"total_observations": total_observations,
"status": camera.get('status', 'unknown') if isinstance(camera, dict) else 'unknown'
}
# 3. Time-based Analytics
# Calculate average dwell time
dwell_times = []
for identity in all_identities.values():
dwell_time = identity.last_seen - identity.first_seen
dwell_times.append(dwell_time)
avg_dwell_time = sum(dwell_times) / len(dwell_times) if dwell_times else 0
# 4. Movement Patterns
cross_camera_movements = 0
for identity in all_identities.values():
if len(identity.cameras_seen) > 1:
cross_camera_movements += 1
# 5. Recent Activity (last hour)
one_hour_ago = current_time - 3600
recent_identities = len([id for id, identity in all_identities.items()
if identity.first_seen > one_hour_ago])
# 6. Detection Quality
avg_confidence = sum([identity.confidence for identity in all_identities.values()]) / total_identities if total_identities > 0 else 0
# 7. System Performance
system_stats = tracking_system.get_system_stats() if tracking_system else {}
return {
"overview": {
"total_identities": total_identities,
"active_identities": active_identities,
"inactive_identities": total_identities - active_identities,
"recent_identities_1h": recent_identities
},
"camera_analytics": camera_stats,
"time_analytics": {
"average_dwell_time_seconds": round(avg_dwell_time, 2),
"average_dwell_time_minutes": round(avg_dwell_time / 60, 2)
},
"movement_analytics": {
"cross_camera_movements": cross_camera_movements,
"percentage_cross_camera": round((cross_camera_movements / total_identities * 100) if total_identities > 0 else 0, 2)
},
"quality_metrics": {
"average_confidence": round(avg_confidence, 3),
"high_confidence_count": len([i for i in all_identities.values() if i.confidence > 0.8]),
"medium_confidence_count": len([i for i in all_identities.values() if 0.5 < i.confidence <= 0.8]),
"low_confidence_count": len([i for i in all_identities.values() if i.confidence <= 0.5])
},
"system_performance": system_stats,
"timestamp": current_time
}
@app.websocket("/api/ws")
async def websocket_endpoint(websocket: WebSocket):
"""WebSocket endpoint for real-time alerts and updates"""
await websocket.accept()
active_connections.append(websocket)
print(f"[WebSocket] Client connected. Total connections: {len(active_connections)}")
# Track which identities we've already alerted for
alerted_identities = set()
try:
while True:
await asyncio.sleep(2) # Check every 2 seconds
# Send alerts for new global identities
if tracking_system:
identities = tracking_system.global_identities
for global_id, identity in identities.items():
# Skip if we've already alerted for this identity
if global_id in alerted_identities:
continue
# Alert for new high-confidence identity
if identity.confidence > 0.75:
alert = {
"type": "alert",
"alert_type": "new_identity",
"message": f"New person detected: {global_id[:12]}",
"timestamp": time.time(),
"data": {
"global_id": global_id,
"confidence": identity.confidence,
"cameras_seen": identity.cameras_seen,
"person_name": global_id[:12]
}
}
try:
await websocket.send_json(alert)
alerted_identities.add(global_id)
print(f"[Alert] Sent new_identity alert for {global_id[:12]}")
except Exception as e:
print(f"[Alert] Error sending alert: {e}")
break
except WebSocketDisconnect:
print(f"[WebSocket] Client disconnected")
if websocket in active_connections:
active_connections.remove(websocket)
except Exception as e:
print(f"[WebSocket] Error: {e}")
if websocket in active_connections:
active_connections.remove(websocket)
@app.on_event("startup")
async def startup_event():
try:
initialize_system()
except Exception as e:
print(f"Failed to initialize Lucid system on startup: {e}")
if __name__ == "__main__":
# Initialize system
try:
initialize_system()
except Exception as e:
print(f"Failed to initialize system: {e}")
exit(1)
# Run server
uvicorn.run(app, host="0.0.0.0", port=8000)